Panbiogeography of the Americas

[Stuart G. Poss]

After reading Croziat, who as others have also noted could have greatly
benefited from the assistance of an editor and peer review, I could never figure
out exactly how tracks are to be drawn (defined, identified) nor have
practioners, to my knoweledge, ever been precise in describing this proceedure.
There has been much discussion of tracks, but these always seem to be described
as some relatively vague association between two or more non-contiguous
distributions of presumtive sister-taxa.

Does one connect each locality collected with all others via a line to form the
track as the work of some authors seems to imply, or is the mean position or
some other geometric or statistical measure of cluster density of a set of
points used to define the track geometry? Does it even need to be a line (arc)?
If not, how much curvature over the surface (wiggle or deviation from a true
arc) is permissible and by what criteria does one decide?  If it is not a line
or curve does it have width?  If so, what criteria are used to establish how
much?  What about disjunct distritibutions?  Does one draw separate tracks for
each segment, when connecting it to that of its presumptive sister-species?  How
are outliers, "odd" vagrants, or potential locality record errors addressed by
this technique when establishing the intial tracks?

How does the technique for track deliniation (?definition) take into account
distributional (latitudinal or positional) shifts in distributions through time
when "drawing" and comparing tracks based upon contemporary distributions?  How
does the technique account for seaonal or temporal variation in range
boundaries?  How does the methodology for establishing  track geometry account
for extinction over a portion of the geographic range?  Obviously, extinction
can result in certain "track geometries" characterizing extant distributions
being simply methodological artifacts unrelated to the historical processes
responsible for producing the observed distributions.  How does one distinguish
"true tracks" from those that are simply the result of chance associations?
There was a paper discussing the inability of track analysis to address last
issue, I believe by Simberloff and colleagues, but I can't remember the exact
citation.  Perhaps someone can refresh my memory.  I do not recall seeing a
presentation rebutting these objections, except for a few vauge pontifications
that they were somehow irrelevant to track analysis.

Some authors, including Croizat, seem to like to use a sort of "inferred or
approximate" convex hull around extant distributions, crudely drawn to describe
(deliniate? define?) tracks. What is the methodological and emperical basis for
use of such a "track geometry" for a given set of species?  Clearly, such a
methodology would be inappropriate in the case where just one or a few
vagrant/imigrant individuals from only a portion of the former range were able
to get from point A to point B, as we now know is often the case for many
non-indigennous species now living in the US and whose presumtive track is a
reflection of the route taken by the airplane or ship bearing the founder(s) and
which could not be inferred by simply looking at the extant distribution (track
geometry) alone (molecular and morphometric analysis are, however, often useful
in confirming that these routes "tracks" were the ones taken and accounting for
the extant distribution).

Given a track, however drawn, how does one distinguigh among the many potential
historical mechanisms that might be drawn with the same track (ie disperal along
the track as in many pan-austral distributions of marine organisms vs.
continental drift [see, as an example, arguments between Rosen and McDowell
regarding distribution of the Galaxiidae)?  How does one distinguish among
mechanisms (tracks?) with the same track track geometry that may be separated in
time?  Given a spheroidal earth, how does one decide, which direction (ie
over/through what intermediate geographic regions) to draw the connection
(association)?

In my opinion Croizat did not give any clear answers to these questions, which
must be in hand before one could begin to conceive of applying "quantative"
methods  necessary for the development of algorithsms for track analysis (ie
definition of track construction methodology precise enough to "explain" to a
machine)..

Perhaps with answers to such questions, the value of Croziatian biogeographic
analysis can be better appreciated.  Are there any references which provide
criteria for precisely establishing track geometry?  Remarkably, these do not
seem to be cited in most works utilizing this approach, but intead often cite
Croizat, who to my knowledge, provided none.

John Grehan wrote:

>  fred schueler wrote:
>
> >> * if your system of connecting lines is entrained in a system
> >> that encodes them as a falsifiable hypothesis, then that sounds like
> >> science to me.
>
> There are techniques proposed by which track relationships may be falsified
> so the method meets this criterion. There is a lot of work to be done in this
> area, but there are few quantitative people who also have a strong interest
> in geography.
>



--
_____________________________________________________________________
Stuart G. Poss                       E-mail: Stuart.Poss at usm.edu
Senior Research Scientist & Curator  Tel: (228)872-4238
Gulf Coast Research Laboratory       FAX: (228)872-4204
P.O. Box 7000
Ocean Springs, MS  39566-7000